The utility of the standardized uptake value, metabolic tumor volume and total lesion glycolysis as predictive markers of recurrent breast cancer

Breast cancer is the second leading cancer killer of women globally. An early measure utilizing a noninvasive molecular marker for predicting cancer aggressiveness is important to better manage the patient and to avert early disease progression. We aimed to determine whether metabolic tumor volume (MTV) and total lesion glycolysis (TLG) are able to predict risk in high TNM tumor staging and the need for the appropriate treatment in breast cancer patients. This is a retrospective study of confirmed breast cancer patients who underwent neoadjuvant, local and adjuvant treatment and follow-up. The 18F-FDG PET/CT study for initial staging was performed, and metabolic parameters (MTV, TLG, SUVmax mean) were analyzed. Spearman correlation was used to assess correlations between metabolic parameters and clinicopathological factors with TNM staging and treatment intention. SUVmean, wbMTV and wbTLG were analyzed to predict the dichotomization of patient endpoint for low (stage I and II) and high (stage III and IV) TNM stage. Twenty-six patients (4 low stage, 22 high stage) with a mean age of 51.8 ± 11.8 years with confirmed breast cancer underwent 18FFDG PET/CT. The MTV and TLG parameters in the tumor (T) were significantly correlated with the TNM stage (P < 0.050); the SUVmax mean (4.18 ± 1.68 g/dl), wbMTV mean (404.68 ± 558.02 cm3) and wbTLG (1756.55 ± 2432.11 g) differed significantly in the high versus low TNM staging with the best predictive cut-off value of SUVmax mean (3.55 g/ml, p < 0.05), wbMTV (20 cm3, p < 0.05) and wbTLG (130 g, p < 0.05) when these values were exceeded. Only wbTLG (130 g, p < 0.05) showed significance difference in treatment intention. In this study, the metabolic parameters SUVmax mean, MTV and TLG showed potential good relationships with TNM staging. TLG was the only marker that influenced the treatment intention in predicting breast cancer aggressiveness.


Background
The global cancer statistics for 2012 show that one out of four of all cancer cases and 15% of all cancer deaths among females are due to breast cancer [1]. Referring to a locally conducted survey, the Malaysian National Cancer been the focus of the medical imaging community in recent years for determining the staging and follow-up of various malignancies, including breast cancer [3,4]. The improved accuracy of TNM staging by PET/CT is due to the ability of conventional imaging modalities to provide information on metabolic activity early at the cellular level [5]. Tissue metabolic activity utilizing the 3D volumetric tumor volume, i.e., wbTLG and wbMTV, can be quantified using 18 F-FDG PET-CT. The quantitative information provided indicates the functional state of tumor cells [6]. To date, multiple risk factors associated with breast cancer have been identified, but the exact etiology remains a mystery. Moreover, no definite preventive strategy has been established to reduce the incidence of breast cancer or cancer-related morbidity and mortality [7]. Therefore, in this study, we aimed to discover the effect of the use of metabolic markers for predicting recurrent breast cancer on TNM staging and treatment outcome [8,9].

Methods
This study was primarily aimed at discovering the usefulness of the metabolic markers of 18 F FDG PET-CT in predicting disease outcomes. The study was approved by the national ethics committee. This was a retrospective study (2018)(2019)(2020) with secondary data analysis. Blinded sampling was carried out on the clinical database of 51 patients who were referred to the endocrinology clinic with a proven histological biopsy of breast cancer. Only 26 patients with follow-up at the clinic for re-staging on 18 F FDG PET-CT who had at least more than 1 year of disease-free progression were included. All patients had confirmed histology for the recurrent tumor. Twenty-five patients that were new cases were excluded. Other exclusion criteria included patients with diabetes mellitus or poor serum glucose control who underwent 18 F FDG PET CT (glucose level of more than 10 mmol/L), very poor renal function (blood urea level of more than 15 mg/dl), concomitant illness such as infection, sepsis or inflammatory process, lack of follow-up data or demise. All patients were followed-up for at most 2 years to evaluate the treatment response and for prognostication (Fig. 1). The subjects were dichotomized into groups regarding TNM stage (high versus low) and treatment intention (yes versus no). The treatment intention was defined as whether the patient had no chemo-radiotherapy with/ without surgery (no) or had chemo-radiotherapy with/ without surgery (yes). Whole-body 18 F-FDG PET/CT scans were performed using a Discovery 610, 16 slice MDCT, with a BGO crystal PET camera. Instructions were given to patients to fast for at least 6 h prior to 18 F-FDG intravenous injection (5.2 MBq/kg). Blood glucose was monitored and was     the CT image data were used to automatically position the patient for PET acquisition. All images were reconstructed using an iterative algorithm and were transferred to a dedicated workstation (SyngoVia) and analyzed using syngo True-D software. Every hypermetabolic lesion was selected, and a volume of interest (VOI) placed manually over the visible area of a target lesion on each PET image (Figs. 1, 2, 3, 4).

Image evaluation
Retrieved images were read by a radiologist with more than 10 years of working exposure. The radiologist was blinded to the previous PET-CT reporting results and the clinical diagnosis of the patients. A positive lesion on PET-CT ( Fig. 1) was determined when the uptake of the lesion in question appeared higher than the FDG uptake in the background (mediastinal blood pool in the thorax or liver uptake in the abdomen) [8,9]. A negative lesion was determined when the suspected lesion had lower FDG-avidity than the background (Fig. 2). Standardization of every PET reading was performed to establish consistent image quality by adjusting the qualitative parameter based on the FDG physiological accumulation in the body (Fig. 3). These parameters were used based on the PET qualitative criteria-Rod's Rule [9]. Positive lesions were also determined as having an SUVmax (g/ dl) of more than 2.5 as a standard for the semiquantitative measurement, and therefore we recorded the mean of SUVmax for all lesions based on the TNM staging (Figs. 1, 2) ( Table 1).

Semiquantitative analysis
The summation of all the lesion volumes of a patient was defined as whole body MTV (wbMTV). The whole body TLG (wbTLG) was calculated as the sum of the TLG values for each lesion in one patient [10] (Fig. 5). All data were transferred for statistical analysis.

Data analysis
Statistical analysis was carried out using Statistical Analysis of Social Sciences System (SPSS) Version 23. Data are presented as the mean and standard deviation (SD) of numerical data such as age, mean SUV, whole-body MTV and whole-body TLG. Spearman correlations were performed to test correlations between nonparametric variables. Cut-off values for mean SUVmax, whole-body TLG and whole-body MTV were determined using receiver operating characteristic (ROC) curve analysis. Treatment intention was used as a dependent variable in the ROC curve analysis. Pearson Chi-squared tests and Fisher's exact tests were performed to test the association between two independent categorical variables. A p value less than 0.05 (< 0.05) was taken to indicate statistical significance.
The dependent variables in this study include the TNM staging of breast cancer, as well as treatment intent. The independent variables in this study include 18 F-FDG PET/ CT derived metabolic parameters such as mean SUV, wbMTV and wbTLG and clinicopathological parameters such as age, menopause status, BMI and obesity status.

Correlation between obesity, menopause and TNM stage
The correlations between obesity, menopause and breast cancer TNM stage of all patients (n = 26) were analyzed. Menopause and obesity were weakly inversely correlated with TNM stage (Fig. 6). However, these parameters were not significantly correlated (p > 0.05).

Correlation of mean SUVmax with TNM stage
There was a strong correlation between mean SUVmax and TNM stage in all patients (n = 26) ( Table 3) (p < 0.05). This means that a high mean SUVmax (4.18 ± 1.68 g/dl) was associated with a high TNM stage (Fig. 7). The mean SUVmax was plotted on a receiver operating characteristic (ROC) curve and analyzed. The area under the curve (AUC) was 0.706, with a 95% confidence interval (CI) of 0.503-0.909 (Fig. 8).

Whole body metabolic tumor volume (wbMTV)
The wbMTV (mean 404.68 cm 3 ) and TNM were strongly and significantly correlated (p < 0.05). This means that a high wbMTV (404.68 ± 558.02 cm 3 ) was associated with a high TNM stage ( Table 5).
As shown in Table 6, there were significantly more patients with a high TNM stage (14, 53.8%) who had a wbMTV of more than 20 cm 3 , compared to the low TNM stage group (8,30.8%) (p < 0.05; Fisher's exact test).

Correlation of wbMTV and treatment intention
wbMTV and treatment intention were poorly correlated in the 26 patients (p > 0.05) ( Table 7).

Correlation of wbTLG with TNM stage
The correlation of wbTLG with the TNM stage of 26 patients is shown in Table 8. The wbTLG was strongly and significantly correlated with TNM stage (p < 0.05). This means that a higher wbTLG value (mean: 1756.55 ± 2432.11 g) was associated with a high TNM stage. However, there was a poor correlation between TNM staging and treatment intent (p > 0.05). The wbTLG was plotted on the receiver operating characteristic (ROC) curve and analyzed. The area under the curve (AUC) was 0.634, with a 95% confidence interval (CI) of 0.411-0.857 (Fig. 7).

wbTLG with treatment intention
There were significantly more patients who underwent treatment (8, 30.8%) who had a wbTLG greater than 130 g, compared to patients without treatment (1, 3.8%) (p < 0.05; Fisher's exact test) ( Table 10). The logistic regression model was used to determine the most potent predictive factors for TNM and treatment intention. The results indicated that there was no association between TLG, interaction between TLG and SUV, interaction between TLG and MTV and interaction between TLG, SUV and MTV with TNM staging.

Discussion
Prediction of the effect of treatment intention by the treating physician on the TNM staging is important in characterizing the cancer aggressiveness of recurrent tumors. The utility of metabolic markers measured by 18 F FDG PET has taken center stage as the parameters underpin the cellular metabolic derangement caused by proliferative mitosis [10]. It is thus useful to assess the effect of prediction of these markers on treatment outcome and its association with TNM staging. In our study, despite the known association of the clinical parameters of age, menopause, BMI, obesity and metabolic parameters, there were no correlations that could have predicted TNM staging more accurately than previous studies [11,12].
In our study, the mean SUVmax (4.18 ± 1.68 g/dl) was strongly correlated with the TNM stage. This finding was consistent with other studies with comparable findings, indicating that the rate of glycolytic uptake of tumor cells   has a strong correlation with higher staging of the disease (stage III & IV) at diagnosis [13][14][15]. SUV is used as a popular predictive metabolic marker for staging and in characterizing tumor cellular aggressiveness. Nevertheless, its utility falls short because of variability in its values in different tumor lines and technical factors [16]. The SUV uptake of 18 F FDG samples the 2D tumor dimension randomly and glucose uptake, which may underestimate the true SUV value of the whole lesion [17]. Therefore, 3D    [10]. Another study involving 40 breast cancer patients also showed that those with an MTV T ≥ 19.3 cm 3 and TLG T ≥ 74.4 g were 10-12 times more likely to recidivate in aggressive tumor cells when these thresholds were exceeded [18]. Correspondingly, the MTV and TLG are potentially promising in ascertaining the association between the tendency for localized tumor invasion, distant metastasis and overall survival in hypopharyngeal cancer [19][20][21][22]. The cut-off values for the mean SUVmax (3.55 g/ml p = 0.033), wbMTV (≥ 20 cm 3 , p = 0.033) and wbTLG (≥ 130 g, p = 0.022) could potentially become good predictors of breast cancer aggressiveness. The values of these parameters are consistent with those found in another study, which reported the best cut-off values of whole-body SUVmax (20.4 g/dL, AUC 0.79, p = 0.011), wbMTV (38.1 cm 3 , AUC 0.81, p = 0.008) and wbTLG (169.1 g, AUC 0.78, p = 0.015) for predicting disease recurrence in a group of locally advanced breast cancer patients [23,24]. A study by An et al. comprising 173 patients with invasive breast carcinoma revealed that the SUVmean (> 1.2, p = 0.033) and MTV (> 2.38 cm 3 , p = 0.005) values of breast tumors are statistically powerful for predicting axillary lymph node metastasis in T1 and T2/T3 breast cancer, respectively [25,26].
On another note, the noninvasive metabolic markers SUV, wbMTV and wbTLG could potentially be used as a surrogate marker to the histological findings. In our study, wbTLG (130 g) was the only potentially predictive  marker for determining the treatment outcome. A higher wbTLG value is likely to predict more patients who may undergo systemic treatment than surgical treatment alone. Few reports in the literature have discussed the reliability of wbTLG, and our study may set a new standard for its use as a metabolic marker in evaluating breast cancer aggressiveness. Our study has a few limitations in relation to the population at large. The number of study subjects was relatively small, which may have led to bias in determining the true values in assessing the metabolic markers as predictors of breast cancer aggressiveness. Furthermore, the study was performed retrospectively; therefore, the incremental value of 18 F FDG PET_CT in evaluating the disease outcome was based on the clinical assessment of the patients' cohort, given that no post-treatment PET imaging was available [26,27]. Finally, only one reader was assigned to read the sampled PET data which was substantiated by the quantitative SUVmax standard reference to ensure the consistency of the results.

Conclusions
In this study, the SUV, wbMTV and wbTLG were found to be potential metabolic markers predicting a high TNM stage and the treatment intention. The 3D assessment of the tumor metabolic volumes by MTV and TLG gave new insight into the reliability of the metabolic parameters to be used as predictive markers. In particular, the cut-off value points to predict the likelihood of patient be subjected to systemic therapy than surgery where wbTLG > 130 g being more likely to recidivate when this threshold were exceeded.